1,577 Hits in 7.5 sec

Robust conditional GANs under missing or uncertain labels [article]

Kiran Koshy Thekumparampil, Sewoong Oh, Ashish Khetan
2019 arXiv   pre-print
We design a new training algorithm, which is robust to missing or ambiguous labels.  ...  labels.  ...  Conclusion We proposed a robust conditional GAN (RCGAN) architecture which was theoretically shown to be robust to a general class of uncertain labels.  ... 
arXiv:1906.03579v1 fatcat:3qozejre5feztkmzka6wajdrhq

Partially Conditioned Generative Adversarial Networks [article]

Francisco J. Ibarrola, Nishant Ravikumar, Alejandro F. Frangi
2020 arXiv   pre-print
With the introduction of Conditional GANs and their variants, these methods were extended to generating samples conditioned on ancillary information available for each sample within the dataset.  ...  the standard approach under these circumstances.  ...  Furthermore, it is immediately clear that the architecture is robust to training under some degree of missing conditioning data.  ... 
arXiv:2007.02845v1 fatcat:raddxtwtxbg5ffbga6zdtcdr6m

Using generative adversarial networks to evaluate robustness of reinforcement learning agents against uncertainties

Fazel Khayatian, Zoltán Nagy, Andrew Bollinger
2021 Energy and Buildings  
As an original contribution, we propose a conditional deep convolutional Generative Adversarial Network (GAN) for projecting multi-dimensional time-series profiles of building performance.  ...  Highlights  GAN is utilized to create synthetic building performance profiles  Synthetic projections are conditioned based on climate and building operation  Uncertainty is infused into synthetic data  ...  Comparing the performance of the RL agent with that of the RBC under uncertain scenarios also shows that the superiority of the RL agent can greatly vary depending on the objective.  ... 
doi:10.1016/j.enbuild.2021.111334 fatcat:imbffqbkardobfs52pg7czzuym

Tag-assisted Multimodal Sentiment Analysis under Uncertain Missing Modalities [article]

Jiandian Zeng, Tianyi Liu, Jiantao Zhou
2022 arXiv   pre-print
To this end, in this paper, we propose a Tag-Assisted Transformer Encoder (TATE) network to handle the problem of missing uncertain modalities.  ...  Multimodal sentiment analysis has been studied under the assumption that all modalities are available.  ...  The primary task is to classify the overall sentiment (positive, neutral, or negative) under uncertain missing modalities.  ... 
arXiv:2204.13707v1 fatcat:rpqm2jypfrhsnliagrk5cgkn7u

RDIS: Random Drop Imputation with Self-Training for Incomplete Time Series Data [article]

Tae-Min Choi, Ji-Su Kang, Jong-Hwan Kim
2020 arXiv   pre-print
It is common that time-series data with missing values are encountered in many fields such as in finance, meteorology, and robotics. Imputation is an intrinsic method to handle such missing values.  ...  Also, self-training is introduced to exploit the original missing values without ground truth.  ...  We evaluated the imputation performance under different missing rates to show that RDIS is robust to various missing rates.  ... 
arXiv:2010.10075v1 fatcat:iosbu7rbpjfnpoaizdijwig77e

Synthesising Electronic Health Records: Cystic Fibrosis Patient Group [article]

Emily Muller, Xu Zheng, Jer Hayes
2022 arXiv   pre-print
Balanced classes can be obtained by oversampling exact copies, with noise, or interpolation between nearest neighbours (as in traditional SMOTE methods).  ...  We remove all demographic or measurement variables with greater than 50% missingness and uncertain encoding, resulting in a binary matrix of 41 variables (see Appendix Table 1 ).  ...  However, conditioning on the outcome class is not always desirable or available in practice.  ... 
arXiv:2201.05400v1 fatcat:vikmozvitbei3efh26lheyinpy

Single-cell RNA-seq Imputation using Generative Adversarial Networks [article]

Yungang Xu, Zhigang Zhang, Lei You, Jiajia Liu, Zhiwei Fan, Xiaobo zhou
2020 bioRxiv   pre-print
ScIGANs is also scalable and robust to small datasets that have few genes with low expression and/or cell-to-cell variance.  ...  Then, a scRNA-seq matrix with cells will be represented as grayscale images and used to train a conditional GANs with the cell labels.  ...  Although GANs is a supervised model that requires pre-defined cell labels, we implemented scIGANs to accommodate scRNA-seq data without prior labels, instead to learn the labels by applying spectral clustering  ... 
doi:10.1101/2020.01.20.913384 fatcat:vzebhtgo2bcptojeivaqan5fwq

Hidden Footprints: Learning Contextual Walkability from 3D Human Trails [article]

Jin Sun, Hadar Averbuch-Elor, Qianqian Wang, Noah Snavely
2020 arXiv   pre-print
We tackle this problem by leveraging information from existing datasets, without additional labeling.  ...  We devise a training strategy designed for such sparse labels, combining a class-balanced classification loss with a contextual adversarial loss.  ...  Specifically, we labeled 50 randomly selected scenes. For each image, we draw a walkability map of where a pedestrian might walk under normal conditions.  ... 
arXiv:2008.08701v1 fatcat:ehqdyoqktfgn5cdxwzf3k7gzky

Deep Learning in Medical Imaging

Mingyu Kim, Jihye Yun, Yongwon Cho, Keewon Shin, Ryoungwoo Jang, Hyun-jin Bae, Namkug Kim
2019 Neurospine  
The advancement of computing power with graphics processing units and the availability of large data acquisition, deep neural network outperforms human or other ML capabilities in computer vision and speech  ...  Machine learning (ML) is a subset of AI that learns data itself with minimum human intervention to classify categories or predict future or uncertain conditions. 1 Since ML is data-driven learning, it  ...  To acquire strong labels for detecting disease patterns or conditions is expensive and inevitable in medical environments.  ... 
doi:10.14245/ns.1938396.198 pmid:31905454 pmcid:PMC6945006 fatcat:miszi3fiojh35ldsgggxgpaowa

scIGANs: single-cell RNA-seq imputation using generative adversarial networks

Yungang Xu, Zhigang Zhang, Lei You, Jiajia Liu, Zhiwei Fan, Xiaobo Zhou
2020 Nucleic Acids Research  
ScIGANs is robust to small datasets that have very few genes with low expression and/or cell-to-cell variance.  ...  We demonstrated in many ways with compelling evidence that scIGANs is not only an application of GANs in omics data but also represents a competing imputation method for the scRNA-seq data.  ...  Then, a scRNAseq matrix with M cells will be represented as M grayscale images and used to train a conditional GANs with the cell labels.  ... 
doi:10.1093/nar/gkaa506 pmid:32588900 fatcat:inqnoipnfnfrfoyuswk5oycvei

A Survey on Deep Semi-supervised Learning [article]

Xiangli Yang, Zixing Song, Irwin King, Zenglin Xu
2021 arXiv   pre-print
We first present a taxonomy for deep semi-supervised learning that categorizes existing methods, including deep generative methods, consistency regularization methods, graph-based methods, pseudo-labeling  ...  Structured GAN. Structured GAN [165] studies the problem of semi-supervised conditional generative modeling based on designated semantics or structures.  ...  Under semi-supervised conditions, when a large number of labels are unobserved, the key to this kind of method is how to deal with the latent variables and label information.  ... 
arXiv:2103.00550v2 fatcat:lymncf5wavgkhaenbvqlyvhuaa

MimicGAN: Corruption-Mimicking for Blind Image Recovery & Adversarial Defense [article]

Rushil Anirudh, Jayaraman J. Thiagarajan, Bhavya Kailkhura, Timo Bremer
2018 arXiv   pre-print
Existing techniques either explicitly construct an inverse mapping using prior knowledge about the corruption, or learn the inverse directly using a large collection of examples.  ...  We present MimicGAN, an unsupervised technique to solve general inverse problems based on image priors in the form of generative adversarial networks (GANs).  ...  , i.e. in case of an unknown sensor or uncertain physics process.  ... 
arXiv:1811.08484v1 fatcat:7yxefwgjh5dj7pybdfzmj2unji

GAN-EM: GAN based EM learning framework [article]

Wentian Zhao, Shaojie Wang, Zhihuai Xie, Jing Shi, Chenliang Xu
2018 arXiv   pre-print
Specifically, a conditional generator captures data distribution for K classes, and a discriminator tells whether a sample is real or fake for each class.  ...  In M-step, we design a novel loss function for discriminator of GAN to perform maximum likelihood estimation (MLE) on data with soft class label assignments.  ...  Generator: Similar to conditional GAN [18] , apart from the random noise z, class label information c is also added to the input of the generator.  ... 
arXiv:1812.00335v1 fatcat:zuiur6wq4vh7tmy7xuinkdh2ni

A Unifying Review of Deep and Shallow Anomaly Detection [article]

Lukas Ruff, Jacob R. Kauffmann, Robert A. Vandermeulen, Grégoire Montavon, Wojciech Samek, Marius Kloft, Thomas G. Dietterich, Klaus-Robert Müller
2020 arXiv   pre-print
Deep learning approaches to anomaly detection have recently improved the state of the art in detection performance on complex datasets such as large collections of images or text.  ...  In particular, we draw connections between classic 'shallow' and novel deep approaches and show how this relation might cross-fertilize or extend both directions.  ...  type I error), and (ii) undetected or missed true anomalies (i.e., the miss rate or type II error).  ... 
arXiv:2009.11732v2 fatcat:4ppfpds3ivd3bk5xcdoxmzmlie

Towards Robust Pattern Recognition: A Review [article]

Xu-Yao Zhang, Cheng-Lin Liu, Ching Y. Suen
2020 arXiv   pre-print
, under weak or noisy supervision.  ...  The accuracies for many pattern recognition tasks have increased rapidly year by year, achieving or even outperforming human performance.  ...  Any disturbances on the number or the labeling condition of the data will usually lead to great changes on the final performance.  ... 
arXiv:2006.06976v1 fatcat:mn35i7bmhngl5hxr3vukdcmmde
« Previous Showing results 1 — 15 out of 1,577 results